Table 10 Performance comparison with different patch sizes.

From: Multi-kernel inception-enhanced vision transformer for plant leaf disease recognition

Patch size

Training loss

Training Acc

Validation loss

Validation Acc

Apple dataset41

 2

0.3741

0.9912

0.3302

0.9874

 4

0.3828

0.9914

0.3722

0.9914

 8

0.3728

0.9972

0.3732

0.9923

 16

0.3875

0.9905

0.3826

0.9841

Rice dataset34

 2

0.3354

0.9942

0.3671

0.9901

 4

0.3249

1.0000

0.3521

0.9942

 8

0.3228

1.0000

0.3528

0.9970

 16

0.3485

0.9912

0.3662

0.9897

Ibean dataset36

 2

0.3334

0.9843

0.4264

0.9579

 4

0.3157

0.9904

0.4178

0.9617

 8

0.3028

0.9965

0.4028

0.9702

 16

0.3394

0.9808

0.4349

0.9496

Cassava dataset42

 2

0.4417

0.8846

0.6579

0.7319

 4

0.4259

0.9155

0.6297

0.7552

 8

0.4228

0.9247

0.6328

0.7651

 16

0.4491

0.8808

0.6592

0.7296

PlantVillage dataset14

 2

0.1507

0.9904

0.2142

0.9902

 4

0.1358

0.9947

0.1857

0.9926

 8

0.1347

0.9973

0.1836

0.9941

 16

0.2268

0.9923

0.1926

0.9917

Back to article page